Muhammad Rashid

Muhammad Rashid

Machine Learning Researcher

Projects 📝🔬

EU Project: DistriMuSe — Safe Interaction with Robots

🔗 https://distrimuse.eu/

Contributed to the Safe Interaction with Robots use case of the European industrial project DistriMuSe, focusing on explainable anomaly detection for industrial robotics.

  • Contributions
    • Developed a zoned VAE–GAN anomaly detection framework for localized and interpretable monitoring.
    • Introduced a robust threshold calibration strategy across 53 anomaly scoring methods.
  • Achievements
    • 99.61% accuracy, 95.1% recall, 90.9% F1-score
    • Real-time inference at ~12.5 FPS
  • Tools & Methods
    • Generative AI, Python, PyTorch, VAE–GANs, CNNs

Improvements in Sampling Strategy for LIME Image Explanations

🔗 https://github.com/rashidrao-pk/lime_stratified

Research project focused on improving the sampling strategy of LIME for image explanations, enhancing stability and faithfulness of explanations for deep vision models.

  • Tools & Methods: Python, TensorFlow, CNNs, Explainable AI (LIME)

Explainable Anomaly Detection — Trust Case Study

🔗 https://github.com/rashidrao-pk/anomaly_detection_trust_case_study

A comprehensive study on trust and interpretability in anomaly detection systems, combining generative models with explanation techniques to support human decision-making.

  • Tools & Methods: Generative AI, Python, TensorFlow, VAE–GANs, CNNs

AI Deployment on Edge Devices

🔗 https://github.com/rashidrao-pk/AI_on_Edge_Devices

Research and implementation of lightweight deep learning models for deployment on resource-constrained edge devices.

  • Tools & Methods: CNNs, Model Quantization, TensorFlow, Raspberry Pi

Previously Developed Freelance Research Projects (2017-2022) 🏭

As a freelancer, applied Computer Vision and Machine Learning projects developed through international freelance collaborations with academic and industrial clients. The work mainly covers medical image analysis (detection, segmentation, and classification from MRI, CT, fundus, and dermoscopic images) and classical as well as deep learning–based vision pipelines, including feature matching, image stitching, fusion, enhancement, and denoising.

Several projects also address intelligent decision systems using neural networks, ensemble learning, evolutionary algorithms, and graph-based models. Implemented primarily in MATLAB and Python, many solutions were delivered as end-to-end systems with graphical user interfaces (GUIs), emphasizing practical deployment, interpretability, and real-world applicability in healthcare, automation, and safety-critical scenarios. Further details about projects are given on the THIS GITHUB REPO. Here are few of examples what clients said about me? can be found at This Link

Some of these contributions are publicly available as open-source resources on the MathWorks File Exchange. Below is a brief summary of selected completed projects.